Training Regimes and Skill Formation in France and Germany: An Analysis of Change Between 1970 and 2010 - Thijs Bol
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Training Regimes and Skill Formation in France and Germany 1113
Training Regimes and Skill Formation in France and Germany
Training Regimes and Skill Formation in France
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
and Germany: An Analysis of Change Between 1970
and 2010
Benjamin Elbers, Department of Sociology, Columbia University
Thijs Bol, Department of Sociology, University of Amsterdam
Thomas A. DiPrete, Department of Sociology, Columbia University
ow do educational systems prepare workers for the labor market? Stratification
H research has often made a distinction between two ideal-types: “qualificational
spaces,” exemplified by Germany with a focus on vocational education, and
“organizational spaces,” exemplified by France with a focus on general education.
However, most studies that investigated this distinction did so by focusing only on
the size of the vocational sector, not on whether graduates with a vocational degree
actually link strongly to the labor market. Moreover, they often studied male work-
ers only, ignoring potential gender differences in how school-to-work linkages are
established. In this paper, we map the change in education–occupation linkage in
France and Germany between 1970 and 2010 using an approach that can distinguish
between changes in rates and changes in the structure of school-to-work linkages.
Surprisingly, we find that the German vocational system in 1970 was not, on average,
substantially more efficient in allocating graduates to specific occupations than the
French system. This finding is a major departure from earlier results, and it shows
that the differences between 1970’s France and Germany, on which the qualificational-
organizational distinction is based, are smaller than previously assumed. Partly, this is
due to the fact that the female labor force was omitted from earlier analyses. We thus
show that ignoring the female workforce has consequences for today’s conception of
skill formation systems, particularly because a large share of educational expansion is
caused by an increase in female enrollment in (higher) education.
Introduction
How do educational systems prepare students for the labor market?1 This
question has been the focus of social scientists and policy-makers for a long
time. In describing how countries establish a link between school and work,
social scientists often rely on ideal-types of skill formation systems. An important
© The Author(s) 2020. Published by Oxford University Press on behalf of the Social Forces 99(3) 1113–1145, March 2021
University of North Carolina at Chapel Hill. All rights reserved. For permissions, doi:10.1093/sf/soaa037
please e-mail: journals.permissions@oup.com. Advance Access publication on 27 May 20201114 Social Forces 99(3)
distinction is made between systems that emphasize vocational training, where
work and study are combined to prepare a student for work in a specific
occupation or set of closely related occupations, and systems that emphasize
general training, where education remains largely school-based and skill acqui-
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
sition is more general. A large body of literature relies on this distinction, for
example in trying to explain cross-national variation in youth unemployment,
occupational mobility, status of the first job, or the length of job search (e.g.,
Breen 2005; Bol and Van de Werfhorst 2013), among other topics. The question
of whether skill formation systems are adequately equipped to deal with rapidly
changing skill requirements in today’s labor markets (e.g., through automation
or globalization) has also sparked policy debates in many countries.
One of earliest and influential studies in the school-to-work literature is
Maurice, Sellier, and Silvestre’s book The Social Foundations of Industrial Power
(1986 MSS hereafter), in which they studied how French and German men
made the transition from school to work. MSS concluded that the French
and German skill formation regimes have radically different approaches in
organizing the match between schools and labor markets. France is described
as an organizational space, where workers are hired based on their level of
general education. Germany, in contrast, is described as a qualificational space,
where jobs are closely matched to workers with specific qualifications that
they obtained within the educational system. In Germany, the linkage between
the educational system and the labor market is tight, resulting in a skill-
formation system where specialized educational credentials predict labor market
standing relatively well. In France, on the other hand, this linkage is weaker,
and educational qualifications hold less information about the labor market
positions of workers: workers with the same educational background end up
working in vastly different occupations. MSS influenced several scholars who
used this distinction as an important component of their theoretical work on
the macrostructure of educational systems (Kerckhoff 1995; Shavit and Müller
1998; Allmendinger 1989; Hall and Soskice 2001).
The extent to which the organizational-qualificational dimension adequately
captures the major differences between educational systems, however, has been
called into question by DiPrete et al. (2017). They introduced a new method
for measuring school-to-work linkages, building on the idea that strong school-
to-work linkages imply that occupations are educationally homogeneous. Sur-
prisingly, the authors found that France and Germany differ less than expected
in terms of how closely educational pathways and occupations are linked.
When comparing similar educational pathways across the two countries (say,
university-educated workers in health), the French skill formation system often
performs equally well in leading graduates to occupations that are a good fit
to their skills [see also (Bol et al. 2019)] as the German qualificational space.
Their analysis illuminated the fact that educational systems can differ both on a
compositional dimension (the distribution of students across educational tracks),
and in terms of their structure of association—the strength of linkage.
This similarity between the French and German skill formation system leads
to a puzzle: if MSS were correct, then one or both countries must have changed.Training Regimes and Skill Formation in France and Germany 1115
This would call into the question the quite static treatment of skill-formation
systems in the literature. A second explanation for the puzzle is the treatment
of gender in MSS. Their results are based only on the male labor force, while
DiPrete et al. (2017) studied the whole labor force. We explicitly study potential
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
gender differences, as generalizations about the two skill-formation systems may
not be warranted when a large part of the labor force is disregarded.
This paper addresses this puzzle and maps how skill formation systems change
by providing an empirical description of the possibly changing patterns of
school-to-work linkages. We built upon earlier work by DiPrete et al. (2017),
but introduce the dimension of time by focusing on the evolution of school-to-
work linkages in France and Germany between 1970 and 2010. This requires
the use of a method that is able to distinguish between compositional and
structural changes, as in the last decades both France and Germany have seen
educational expansion and an increase in the labor market participation of
women—processes that are likely associated with changes in school-to-work
linkages.
Country-level ideal types
The study of skill formation systems has usually focused on country-level
differences. Studies have classified educational systems along a number of
dimensions (Allmendinger 1989), and this work was often influenced by MSS’s
qualificational-organizational distinction. A key aspect of this distinction is the
focus on vocational specificity. In their comparative project, Shavit and Müller
(1998) studied rates of vocational education (among other system characteris-
tics) and their implications for the pathways that lead from school to work. Their
key argument is that cross-national variation in how educational systems prepare
students for the labor market is largely based on how many students are enrolled
in vocational education:
“[W]e” distinguish between two ideal-typical regimes of school-to-work tran-
sitions, which, following Maurice, Sellier, and Silvestre, we label qualificational
and organizational spaces. The qualificational space is characterized by a high
rate of specific vocational education. More precisely, a large proportion of
the graduating cohorts leave the educational system with specific skills and
occupational identities. This is in contrast with organizational spaces where
education is predominantly academic or general, and where occupational skills
are learnt on the job or in courses taken after leaving school” (Müller and Shavit
1998, 9).
The consensus in the literature is that school-to-work linkages are stronger
in skill-formation systems that rely on occupation-specific vocational education
and fit the “qualificational space” ideal-type (e.g., Wolbers 2007; Breen 2005;
Müller and Gangl 2003). Students leave the educational system well-prepared
and find work in an appropriate occupation relatively quickly, often because they
enter the labor market with relevant work experience. This also has consequences1116 Social Forces 99(3)
for career mobility, which is higher in organizational spaces (Haller et al. 1985;
König and Müller 1986).
While this literature has established the importance of skill-formation systems
for later-life outcomes, we argue that the identification of countries with specific
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
ideal-typical skill-formation systems rests on assumptions that have rarely been
tested empirically.
First, the approach implies the possibility of distinguishing vocational educa-
tional programs that provide specific skills from academic programs that provide
general skills other through a credentialing requirement to work in specific
occupations, as is typically the case with occupations that are called professions.
However, it is rarely discussed how the skill-content of educational programs
can be measured. For instance, the UNESCO manual for the International
Standard Classification of Education (ISCED) says virtually nothing about how
skill content of vocational education is measured (OECD, UNESCO Institute for
Statistics, European Union 2015, 65).
Second, even if it were possible to distinguish vocational and general pro-
grams, it would be an open empirical question whether specific skills are in fact
associated with stronger school-to-work linkages. There is an implicit assump-
tion that all vocational programs channel graduates to specific occupations, and
so if there are more vocational graduates, the system will be more “occupation-
specific.” This focus on rates of vocational education asserts a substantive focus
on the country as the unit of analysis: rates are computed for the country as
a whole; some countries are said to have vocational educational systems (Ger-
many), and others are not (France or the United States). Given the considerable
variation that is observed between different educational programs within the
same country, we argue that a more fine-grained study of skill-formation systems
is required.
A structural approach
We study the strength of school-to-work pathways as well as cross-national
differences in their composition. We call this “linkage approach” a structural
approach because (1) it does not rely on classifying programs as vocational or
general, and (2) it measures the “success” of educational programs empirically
by studying how closely a program is linked to specific occupations. The
actual content of educational programs is difficult to ascertain, but a central
characteristic of a “successful” vocational program is that it leads to one or
a small set of specific occupations. This characteristic is jointly determined
by educational content (e.g., length and intensity of the program, school-
based vs. work-based programs) and by legal, cultural, and social character-
istics of labor markets (e.g., occupational closure via licensing or collective
bargaining agreements). We argue that a continuous, data-based measure of
the strength of education–occupation linkage is more illuminating and empir-
ically plausible than a binary, theoretically motivated classification (vocational/
academic).
When comparing linkage across countries or within countries over time, a
complication arises. Consider that someone who studied medicine in universityTraining Regimes and Skill Formation in France and Germany 1117
is linked to the labor market asstrongly as agraduate of a car mechanic voca-
tional program. Assume further that the countries differ in their educational
composition: while both countries have the same proportion of car mechanic
graduates, the first country has a higher proportion of medicine graduates. The
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
impact on total linkage from these two educational categories will then be
higher in the country with the higher proportion of medicine graduates. The
difference between the two countries will be purely an effect of the different
marginal distributions. On the other hand, the marginal distributions could be
identical, but the linkage scores for different occupations could be different (say,
in one country medicine graduates link strongly because they are almost all
doctors, while in another country they are sometimes doctors and sometimes
administrators or managers in hospitals). Thus, when comparing countries or
the same country over time, there are always two potential sources of difference:
differences in the marginal distributions of education and occupation, and
differences in the patterns of association.
Differences in marginal distributions are important, but the “success” of
a country’s skill-formation system lies in the effectiveness with which it can
link educational systems and labor markets. Even if many German students
are enrolled in vocational programs, this fact tells us very little about the
extent to which they obtain occupation-specific education. Similarly, even if
in France a much smaller proportion of workers takes a vocational degree,
the extent to which that vocational degree channels school-leavers to the same
occupation might be high. We thus talk about the strength of the association
between specific educational outcomes and the occupational structure, and
we use the term “structural linkage” for this component of total linkage.
The linkage approach provides a clear contrast with existing studies that
focus attention on compositional differences only, i.e., rates of vocational
education.
Patterns of change
An important question in the literature on skill-formation systems is how they
change. Neo-institutionalism has argued that national educational systems were
becoming increasingly standardized and focused on providing general education
(Benavot 1983). The empirical evidence for this claim was the decline in the
rate of secondary vocational education. The comparative project of Müller and
Shavit (1998) came to a different conclusion. To be sure, the authors of the
country-specific analyses saw change taking place to a greater or lesser extent
in the countries they studied: for instance, France is described as a country
that went through a dramatic expansion of secondary schools in the 1950s
and 1960s (Goux and Maurin 1998), while Germany is described as a country
with very stable educational institutions (Müller, Steinmann, and Ell 1998).
However, these changes do not point universally in the direction of a decline
of vocational education, refuting the neo-institutionalists’ view of convergence.
On the contrary, Müller and Shavit (1998) found that skill-formation systems
are relatively ‘fixed’ in time.1118 Social Forces 99(3)
In this paper, we revisit this question of change. We study both change in
the educational and occupational marginal distributions (what we call compo-
sitional difference or change) as well as change in the structure of association
between educational programs and occupations (what we call structural dif-
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
ference or change). We have several reasons for suspecting that not only the
marginal distributions changed in important ways over recent decades, but also
that the structure of association may have changed.
First, DiPrete et al. (2017) have raised questions about the current conception
of French–German educational differences. France at present has a smaller
proportion of workers who were educated in programs that have tight linkage
to the occupational distribution than does Germany. At the same time, many of
France’s educational programs have linkage that is as strong or stronger than in
Germany. The differences between Germany and France, in other words, appear
to involve both structural and compositional differences. However the study by
DiPrete et al. (2017) did not involve historical data. It may thus be the case that
either country has changed in the recent decades, and that the differences were
more pronounced in the 1970s.
Second, from the contributions in Müller and Shavit (1998) and many other
works studying the transformation of western economies, we know that both the
educational and the occupational distribution experienced considerable change
in recent decades. In most countries, higher education has seen an enormous
expansion, and many countries also saw an increase in vocational education.
At the same time, economies have shifted toward the service sector, often
accompanied by deindustrialization and the privatization of formerly state-
owned companies. Educational expansion has likely increased school-to-work
linkages in the aggregate, but the increasing supply of highly educated workers
may have led to a decrease in the structural part of linkage via processes related
to “overeducation.” Similarly, the shift toward the service sector, where jobs are
potentially more loosely defined in terms of their educational requirements, may
have also contributed to structural declines in linkage.
Lastly, two important but rarely discussed aspects of MSS’s study are that it
focused only on large industrial firms and only on the male labor force. This
latter point is also true for follow-up studies (e.g., Haller et al. 1985; König and
Müller 1986). The point about gender was argued by Marry et al. (1998):
The majority of these studies have only been concerned with men [...],
and one could ask whether the label “societal” is really appropriate for
characteristics measured only for the masculine half of society. (p. 356, own
translation).
Given the increase in female labor force participation rates since the 1970s,
this critique has become even more important. In our analyses, we thus pay
special attention to gender differences in the articulation of school-to-work
linkages. The ideal-typical distinction that is still dominant today has been
based only on the male labor force, and so we do not know to what extentTraining Regimes and Skill Formation in France and Germany 1119
“organizational” and “qualificational” spaces also exist for the full labor
force.
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
Analytical strategy
A measure of school-to-work linkage
Our interest lies in the strength of the association between the educational
system and the labor market as a function of time, operationalized here by
the educational level and the field of study on the one hand (using the ISCED
scheme), and current main occupations (ISCO-88) on the other hand. We use
“field of study” here in a broad sense, referring to all education that is specific
to an area of work, including vocational schooling, university degrees, but also
the German Ausbildung in the dual system. The main analytical unit is the
combination of educational level and field of study (“level-field”).
To measure the strength of the link between education and occupation, we
employ a multigroup segregation measure, as introduced by DiPrete et al. (2017).
The link between the educational system and the labor market is tighter when
educational qualifications predict the occupations of workers well. When school-
to-work linkages are low, educational qualifications provide no information
about the occupations of workers. To capture this idea, we use the Mutual
Information Index M (Mora and Ruiz-Castillo 2011; Theil and Finizza 1971).
We define G as the set of combinations of E ISCED levels and F fields of
study. G are thus the “level-fields,” which reflect the specialized educational
qualifications that are available. To calculate M, consider a G × J contingency
table, where each cell counts the number of workers that have one of level-
field G and one of occupation J. To simplify the notation, we transform
the contingency table from a table of counts to a table of proportions (i.e.,
estimates of probabilities). Each cell’s pjg represents the proportion
of workers
in occupation j who were educated in level-field g, and therefore j g pjg = 1.
We also define the marginal probabilities p·g = j pjg and pj· = g pjg , which
simply reflect the overall level-field and occupational distributions. We define the
conditional probability
of being in occupation j, given a level-field g, as pj|g .We
define M G; J as “total linkage,” which measures the dependency between the
level-fields and occupations contained within G and J. Linkage is high for a
specific level-field when the occupational distribution of that level-field deviates
strongly from the overall occupational distribution. We call this the local linkage
L of level-field combination g, and define it as follows:
pj|g
Lg = pj|g ln . (1)
pj·
j
Local linkage will be minimized at zero when pj|g = pj· for all j. In other words,
a level-field is not linked to the labor market if the occupational distribution
for workers that have this level-field is identical to the overall occupational1120 Social Forces 99(3)
distribution. The more strongly the occupational distribution for a level-field
deviates from the overall occupational distribution, the higher the local linkage
for this level-field. The possible maximum value for local linkage depends on
the overall size of the smallest occupation. For instance, assume that the smallest
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
occupation comprises 1 percent of the labor force. If all workers with level-field
g worked in only this occupation (i.e. pj|g is 1 for one j, but 0 for all other j), then
1
local linkage would be maximized at ln .01 ≈ 4.6.
To characterize overall linkage, M is defined simply as the weighted average
of the local linkage scores for all level-fields G:
M G; J = p·g Lg . (2)
g
This additive decomposition of M is helpful in determining where in the
educational distribution the linkage strength of a country originates. Further
details can be found in the more extensive treatment by DiPrete et al. (2017).
Studying change
The M is not a margin-free measure of segregation, i.e., segregation may increase
or decrease depending on changes in the marginal distribution of level-fields
or occupations. This is immediately apparent from equations (1) and (2). This
means that changes in M over time can arise through different mechanisms: either
because of a change in the margins without any structural changes, or through a
change in the structural association between level-fields and occupations. Given
that we are interested in studying differences over time and across countries, we
seek to isolate marginal and structural changes.
We adopt a procedure first proposed by Karmel and Maclachlan (1988) and
extended by Elbers (2020). This procedure is based on the insight that the odds
ratios of the contingency table are the only measure of association that does
not change when the margins are transformed. Consider two G × J contingency
matrices for the same country at different points in time, t1 and t2 . To make a
margin-free comparison, we adjust the margins of the contingency matrix at t1
to be identical to those at t2 . This is achieved by the use of iterative proportional
fitting (IPF), where first the row margins of t1 are scaled toward those of t2 ,
and then the column marginals of the resulting table are scaled toward those
of t2 . Repeating this process several times will transform the marginals of t1 to
approach the marginals of t2 , while preserving the odds ratios of the matrix as
of time t1 . The process is repeated until the marginals are within 0.001 percent
of the marginals at t2 . We call the resulting, counterfactual matrix t1 . We then
repeat the IPF procedure, starting from matrix t2 , to arrive at matrix t2 . These
two scenarios differ in the choice of the original matrix: One adjusts from t1
forward toward t2 , while the other adjusts in a backward direction from t2 to
t1 .Given the four matrices, we calculate M (t1 ) and M (t2 ) as the observed linkage
at time t1 and t2 . We also compute M t1 , as the adjusted t1 linkage, which
can be regarded as the counterfactual linkage at time t2 if only the marginsTraining Regimes and Skill Formation in France and Germany 1121
had changed to equal their values at time t2 . Because M (t2 ) and M (t1 ) have
the same marginal distributions, they differ only in their association structure
(i.e., the odds ratios). The same goes for M (t1 ) and M (t2 ). To arrive at a single
estimate, we average the forward and backward scenarios (often called a Shapley
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
decomposition):
Δmarginal
1 1
M (t2 ) − M (t1 ) = (M (t2 ) − M (t2 )) + (M (t1 ) − M (t1 ))
2 2 (3)
1 1
+ (M (t2 ) − M (t1 )) + (M (t2 ) − M (t1 ))
2 2
Δstructural
Another benefit of this approach is that it allows a further decomposition
of the structural and the marginal component. The marginal change can be
subdivided into two components: one component quantifies the contribution of
changing educational marginals and one quantifies the contribution of changing
occupational marginals. The structural component can be decomposed into the
contributions of each individual level-field. We make use of this property by
summing the contributions of the level-fields using ISCED categories and fields
of study. Details are provided in Appendix A.
The method described here can be used to compare any two M measures,
where “t1 ” and “t2 ” can stand either for different points in time or for different
countries at the same time.
Data
The European Labor Force Survey [EU-LFS, (Eurostat, 2020)] is well suited for
our purposes because it provides harmonized variables for educational levels,
fields of study, and occupations. From 2005 to 2010 datasets, educational levels
and fields of study are coded in the ISCED-1997 scheme, and occupations
are coded using ISCO-88 (3 digits). In 2011, the EU-LFS switched to ISCO-
08, which is why we use only the years up until 2010. Because of the small
sample size in Germany, we pool the samples for 2006–2007 and for 2008–
2009. In the EU-LFS, no fields of study were recorded before 2005. For France,
we additionally use the “Formation et Qualification Professionnelle” [FQP,
(INSEE/ADISP-CMH, 2020b)] survey for 1970 and 1985, as well as the 1990–
2002 series of the French Labor Force Survey [Enquête Emploi, (INSEE/ADIS-
P-CMH, 2020a)]. Here, fields of study are recorded consistently beginning in
1995. To increase the sample size, we pooled the years 1995–1997, 1998–2000,
and 2001–2002. For West Germany, we use the Public Use Files of the 1970 and
1987 censuses (RDC of the Federal Statistical Office and Statistical Offices of the
Länder, 2020).2 We restricted all samples to the current active workforce aged
15 to 643 , leaving out students and the unemployed.
Educational level, field of study, and current main occupation were recorded
in different schemes in many of these years. For educational levels, we used1122 Social Forces 99(3)
Table 1. ISCED-1997 and Native Degrees
Category France Germany
ISCED 1 or less Elementary education or Elementary education or
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
less less
ISCED 2 Brevet, BEPC or some Hauptschul-
secondary education /Realschulabschluss
ISCED 3c CAP, BEP, BP
ISCED 3ab_voc Baccalauréat Lehrabschluss, short
professionnel vocational school
Baccalauréat
technologique
ISCED 3ab_gen Baccalauréat général Abitur,
Fachhochschulreife
ISCED 4 Abitur/Fachhochschulreife
and Lehrabschluss
ISCED 5b DUT, BTS infirmier, Meister, Techniker, long
assistante sociale vocational school
(Fachschule, e.g., in
health)
ISCED 5a/6 DEUG, License, Maîtrise, BA, MA, Diplom,
Diplôme, etc. DEA, Magister, etc. Promotion
Diplôme de docteur
Note: Adapted from official ISCED mappings for France and Germany (http://uis.unesco.org/en/
isced-mappings)
the official ISCED mappings to code degrees into the ISCED-1997 scheme (see
Table 1). The EU-LFS does not provide a breakdown for category 3ab, which
combines workers with general and specialized education. We therefore split this
category using the field of study information: We code a worker as 3ab_voc if a
worker completed a professional or technological baccalauréat (maturity exam)
in France or an Ausbildung (dual training) in Germany; we code as 3ab_gen for
the general baccalauréat in France and the Abitur (maturity exam) in Germany.
We merge 5a and 6 (PhD) because the latter is a small category and cannot be
distinguished in earlier surveys. We also merge ISCED 0 and 1, because these
two categories cannot be distinguished in earlier years. In categories 1, 2 and
3ab_gen, workers have obtained general education, while in all other categories
workers have obtained specialized education (indicated by the presence of a field
of study), be it either through dual training, vocational schooling, or in higher
education. Programs classified as 5b are advanced vocational programs on the
verge of tertiary education, while programs in 3c and 3ab_voc are more often a
mix of on-the-job training and school-based training.
Fields of study have been manually coded into ISCED fields of study (see
Table 2). The crosswalks are found in Online Appendix 1. Except for the FQP
survey in 1970, we were able to use proportional crosswalks to harmonize nativeTraining Regimes and Skill Formation in France and Germany 1123
occupational codings into ISCO-88 [one of these crosswalks is constructed using
data provided by (Bundesinstitut für Berufsbildung [BIBB] 2006)]. To create
a proportional crosswalk, we identified surveys where the native occupational
scheme and ISCO-88 were coded for the same individuals. For each native code,
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
we then calculated the proportion of double-coded ISCO-88 codes. For instance,
the French PCS-1982 code 3751 (Cadres de l’hôtellerie et de la restauration) is
coded in ISCO-88 as 122 (Production and operations department managers)
in 58 percent of the cases, and 131 (General managers) in 42 percent of the
cases. When we apply the crosswalk to our data, we randomly choose with
a probability of 0.58 the ISCO code 122, and 131 otherwise. This process
introduces uncertainties into our estimates, which are in practice very small (see
Online Appendix 2).
Results
Descriptives
Table 2 provides descriptive statistics by year. Percentages are reported for gender
and age groups, as well as for our three main variables of interest.
The tables confirm the well-known patterns of the changing demographics of
labor force participation. Women’s labor force participation increased from 37
percent to 48 percent in France and from 36 percent to 46 percent in Germany—
always slightly below the French levels. The changes in the age structure reflect
both an aging population and an increase in educational attainment. Many of
those aged 15–24 in 1970 were already in the labor force, while in 2010 many
in this age group are still in vocational or tertiary education, entering the labor
market at older ages.
Of special interest are the dramatic changes in the educational structure. In
Table 2, the effects of educational expansion are clearly visible, with increases in
vocational (ISCED 3ab_voc, 3c, and 4) and tertiary education and strong declines
in the share of workers with a “general” field, i.e. ISCED levels 1, 2, and 3ab_gen.
For both countries most of the expansion of tertiary education was in higher
tertiary education (university and post-graduate degrees, ISCED 5a/6) instead of
lower tertiary courses (ISCED 5b).
If we maintained a focus on rates of vocational education, then clearly
the French skill-formation system has changed enormously: for instance, the
combined share of graduates in 3c and 3ab_voc has risen from 17 percent to
43 percent. The numbers also show clear differences in the two countries’ skill-
formation systems. In 1970, Germany roughly 60 percent of the labor force had
undergone vocational education (ISCED 3ab_voc, 4, and 5b), compared to less
than 20 percent in France. In 2010, France’s educational distribution resembles
Germany’s distribution in 1970. The relative shares of both vocational and
tertiary education in France have increased enormously, leading to a labor force
in 2010 with over three quarters of the workers having some sort of specialized
education, up from less than one quarter in 1970. This underscores the stabilityTable 2. Descriptive Statistics by Year
Germany France
1124 Social Forces 99(3)
Census EU-LFS FQP Enq. Emploi EU-LFS
1970 1987 2005 2010 1970 1985 1996 1999 2001 2005 2010
Sample size (in 1000) 1146 1261 185 20 27 28 67 67 45 54 77
Gender
Female 36 38 45 46 39 42 44 45 45 46 47
Age
15–24 17 15 6 6 21 13 5 5 6 7 7
25–34 26 26 20 20 21 30 28 28 26 26 25
35–44 23 24 33 28 24 27 31 30 30 29 29
45–54 18 25 27 30 20 20 27 29 29 27 27
55–64 15 10 13 16 14 10 9 8 10 10 12
Educational levels (ISCED)
ISCED 1 0 1 2 2 39 20 13 10 9 9 6
ISCED 2 32 22 11 8 34 32 22 21 20 18 16
ISCED 3c 16 28 31 31 30 30 27
ISCED 3ab_voc 52 52 49 50 1 3 6 7 7 14 16
ISCED 3ab_gen 1 2 4 2 3 6 6 6 7 1 1
ISCED 4 1 2 7 8
(Continued)
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021Table 2. Continued
Germany France
Census EU-LFS FQP Enq. Emploi EU-LFS
1970 1987 2005 2010 1970 1985 1996 1999 2001 2005 2010
ISCED 5b 7 8 11 10 2 6 11 12 14 12 14
ISCED 5a/6 7 13 17 19 3 6 11 12 13 17 20
Share of general education only
General/No field 33 25 17 13 77 58 41 38 36 28 24
Educational fields (ISCED)
Teacher training, education 2 4 4 4 0 0 0 0 0 1 1
Humanities, languages, arts 2 2 4 4 6 5 4 4 5 8 8
Social sciences, business, law 29 30 30 30 24 28 32 33 33 32 34
Science, maths, computing 2 3 3 3 10 8 3 4 4 8 8
Engineering, manufacturing 50 46 36 34 42 38 38 37 35 31 29
Agriculture, veterinary 4 3 3 3 5 5 5 4 5 4 4
Health, welfare 6 8 13 14 6 11 11 11 11 10 10
Services 5 5 8 8 6 5 7 7 7 5 5
(Continued)
Training Regimes and Skill Formation in France and Germany 1125
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021Table 2. Continued
1126 Social Forces 99(3)
Germany France
Census EU-LFS FQP Enq. Emploi EU-LFS
1970 1987 2005 2010 1970 1985 1996 1999 2001 2005 2010
Occupation (ISCO-88, 1 digit)
Armed forces 1 1 1 1 1 1 1 1 1 1 1
Managers/Senior Officials 3 5 6 6 8 7 8 8 8 8 9
Professionals 7 11 15 16 6 8 10 10 11 13 14
Technicians/Assoc. Profess. 17 21 22 22 10 15 17 17 18 18 19
Clerks 11 13 12 12 12 15 15 15 14 12 11
Service/shop workers 9 10 12 12 6 9 12 12 12 12 13
Skilled agricultural workers 5 3 2 2 14 8 5 4 4 4 3
Craft and related workers 25 20 15 14 20 16 14 14 13 12 11
Plant operators/assemblers 10 8 8 7 10 13 11 11 11 9 9
Elementary occupations 11 10 8 8 13 8 8 8 8 10 10
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021Training Regimes and Skill Formation in France and Germany 1127
of the German skill formation system compared to a number of educational
reforms in France that aimed to increase vocational and tertiary education levels
(Day 2001; Brauns et al. 1999). This pattern of stability in Germany and change
in France is a recurring theme in this paper.
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
Vocational education is traditionally focused on manufacturing and business
degrees, which in both countries constitute the majority of degrees awarded.
With deindustrialization and an increasing focus on service occupations, the
relative share of manufacturing degrees has decreased markedly over time in
both countries. The increasing diversity of vocational education is reflected
by an increase in health, welfare, and services degrees. When comparing the
distribution of fields of study between the countries, they show a roughly similar
pattern.
The distribution of occupational major groups and the patterns of occu-
pational change are similar in both countries, suggesting changes in the skill
distribution of the labor force that are common to both economies. The growth
has been concentrated in high- and medium-skilled occupations (groups 1–5),
while lower-skilled occupations have declined (groups 6–9). There are also some
important differences between France and Germany that seem stable over time.
In both 1970 and 2010, Germany had more workers in craft and professional
occupations (groups 2, 3, and 7), while France had more workers in agriculture,
low-skilled, and management occupations (groups 1, 6, and 9). This pattern
is consistent with MSS’s findings about the organization of work in the two
countries. In France, the number of low-skilled workers and managers is higher,
while Germany relies more on specialized, “medium-skilled” workers.
Observed total linkage over time
Figure 1 shows the strength of linkage for France and Germany both for the
whole labor force, as well as for men and women separately. Each panel also
contains a breakdown by younger (ages 15–34) and older workers (ages 50–
64). The M is measured for each year and gender–age–country combination
separately and defined by applying equation (2) to the matrix of level-fields and
three-digit ISCO-88 occupations.
The absolute value of the M is not strictly interpretable, as the values it can
take depend on the dimensions of the contingency table. While an absolute
interpretation is not possible, by harmonizing the data, we can compare the
M scores in terms of their relative differences—countries over time, between
countries over time, or between genders. For this reason, we will interpret the
changes in M as percentage differences.
Concentrating first on the whole labor force, Germany’s 2010 M was about
38 percent higher than France’s. This is similar to the difference reported in
DiPrete et al. (2017), who used different datasets. In 1970, Germany’s M was
50 percent higher than France’s. Since 1970, Germany’s overall M has increased
by 56 percent,4 and France’s M by 70 percent, with most of this change occurring
before 2000. In terms of total linkage, the countries have thus slightly converged
over time.1128 Social Forces 99(3)
Figure 1. Linkage over time, by gender, age, and country
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
The gender breakdown reveals that in both countries, the rate of linkage
change has been similar for men and women. However, in Germany, men have
slightly higher linkage than women, while in France the opposite is true. One
reason for this could be the different gender-specific occupational composition
of the French and German labor markets. A large share of the German male
workforce is employed in the skilled trades and crafts, which tend to be well-
linked. In France, in contrast, men are more often employed in clerical and service
occupations that tend to be linked less strongly.5
A more striking pattern emerges when one takes into account the differences
between age groups. Figure 1 shows that age differences in linkage are small
for German men, while they are much larger for German women. Age group
differences are larger for men in France than in Germany, but in France (just
as in Germany), the age differences in linkage strength are stronger for women
than for men. To put it differently, older women are in occupations that link
less well than is the case for younger women in both countries, and this age
gap is bigger for women than for men. However, the two countries differ in the
pattern of change. In France, a visible convergence of linkage strength for older
and younger workers has occurred in recent years for both men and women. In
contrast, the linkage-strength gap between older and younger German women
has not closed over the 40 years covered by our data.
Finally, while the size of the age gap varies by country and gender, the relative
pattern is similar and in the expected direction: Younger workers who have justTraining Regimes and Skill Formation in France and Germany 1129
begun their careers link more strongly than older workers with notably stronger
age effects for women than for men. The small differences in linkage strength
for older and younger German men suggests a relative stability of educational
institutions across cohorts and a relatively minor impact of career mobility for
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
the linkage strength of German male workers. In France, the temporal gains
in linkage strength for younger workers may signal that educational reforms
have shifted educational attainment toward more strongly-linking educational
degrees. The even stronger upward shift for older male French workers may
reflect the lagged effects of shifts for younger workers combined with career
mobility patterns that produce higher linkage strength. The especially strong
temporal gain in linkage strength for older female French workers may reflect
changing implications of motherhood on their composition and occupational
placement as well as lagged early career effects.
A pattern that is not easy to distill from Figure 1 is that cross-national
variation in linkage highly depends on which gender one focuses on. Figure 2
sheds light on this issue and shows the difference in linkage between France
and Germany by gender. When only taking men into account (Figure 2, left
panel), our results seem to support MSS’s main conclusions: Germany provides
a tighter school-to-work linkage that France. This conclusion is starkly different
for women (right panel). Especially during the 1970s (the likely period of MSS’s
research), linkage for women is only slightly higher in Germany than it is in
France and the similarities between the two countries are more striking than
the differences. Combined with the results from Figure 1, our findings question
a clear separation between the German and French skill formation systems.
Are they indeed the two extremes of the ideal-typical distinction between
qualificational and organizational spaces, or is this conclusion by MSS only
applicable to part of the labor force?
Next, we study where the linkage strength in Germany and France originates.
The full decomposition of local linkage scores gives 35 terms, one for each
level-field. For a more parsimonious presentation, we sum the contribution
of level-fields to the total M by simplified ISCED levels, where we group
tertiary education (ISCED 5a/6), upper vocational/lower tertiary education (5b),
vocational education (3ab_voc, 3c, 4), and general education (1, 2, 3ab_gen).
The four components are plotted separately by country-years in Figure 3. The
percentage indicates the relative contribution toward total linkage strength, and
it is instructive to compare this number to the proportion of the respective level
among the labor force from Table 2.
In Germany, almost half of the total linkage strength in 1970 originates from
upper secondary vocational education (ISCED levels 3ab_voc/3c/4). Although
only 7 percent of graduates had obtained tertiary education (5a/6), this compo-
nent accounts for 27 percent of total linkage strength. General education (i.e.,
ISCED 1/2/3ab_gen) comprised one-third of Germany’s labor force in 1970, but
contributes only 13 percent to total linkage. In 2010, these relative contributions
have changed only little, and seem to be mostly explained by changes in the
relative proportions. France in 1970 was dominated by general education: 78
percent of the labor force had a general degree, and this group contributes1130 Social Forces 99(3)
Figure 2. Gender-specific patterns of linkage
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
Figure 3. Decomposition of M by ISCED. Note: ISCED 3/4 refers to ISCED 3ab_voc, 3c, and 4
34 percent toward total linkage. While the other three components together
constitute only a small share of the labor force, they contribute a considerable
amount. Two-thirds of France’s linkage in 1970 originates from just 22 percent
of the labor force. France’s changing educational distribution is also reflected in
a different pattern of linkage: In 2010, general education contributes much less
to linkage, and the contributions of ISCED 5 and 6 have grown considerably.Training Regimes and Skill Formation in France and Germany 1131
Figure 4. Decomposition of M between countries (France minus Germany). Note: These results
are also available as a table in Appendix B
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
Between-country differences: then and now
As we noted above, stratification research has often equated high rates of voca-
tional training with high vocational specificity and treated this as a structural
feature of the country’s educational system. Figure 3 already shows that this
conclusion might not be justified: in 1970, France vocational graduates were
only a small share of the labor force, but this group linked strongly to the labor
market. Thus, the internal structure of school-to-work linkages in France and
Germany does not match the conventional wisdom about these two countries.
The findings presented so far have to be interpreted with caution, as they do
not fully account for differences in the educational and occupational marginal
distributions. Comparisons between countries and over time can be mislead-
ing when approached through Figure 3, because they combine structural and
marginal differences. We now present decompositions that allow us to disentan-
gle the possible sources of difference.
First, we apply equation (3) to the differences between countries, both in 1970
and in 20106 (t1 stands for Germany, and t2 for France). The results are shown
graphically in Figure 4, where the left panel decomposes the differences for the
complete labor force, and the right panel contains results separately by gender.
A negative score means that the French component is lower than the German
component.
The top panel shows the raw difference in linkage between the two
countries as observed in Figure 1. This total difference is decomposed into the1132 Social Forces 99(3)
three components shown in the second panel from the top (titled “Overall”):
“occupational margins,” “educational margins,” and “structural.” As we have
described above, the structural change can be further decomposed into the
contributions of the 35 level-fields. For an interpretable presentation of the
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
results, we chose to sum those 35 contributions once by level, and once by field.
These two decompositions are presented in the third and fourth panel from the
top, titled “Structural: Levels” and “Structural: Fields.”
Focusing first on the overall decomposition, France’s linkage was lower than
Germany’s in 1970, but this difference is almost completely accounted for by
differences in the educational marginal contributions. This means that the stark
differences between Germany and France that we observe are mostly a conse-
quence of the differences in educational distributions: In 1970, Germany has
many more workers with vocational education than France. Because vocational
education links more strongly, Germany’s overall linkage is also high. While there
are compositional differences, in 1970 the French educational system was not
less successful than the German system in providing a close link toward the labor
market—it just did so for a smaller share of the labor force. This is the first key
result that emerges from the decomposition.
When comparing separately by gender the pattern is similar. A large negative
component for the educational margins accounts for a majority of the difference
in 1970 for both men and women. What the breakdown by gender makes clear is
that France had a small advantage in terms of the structural component: once we
account for the compositional differences, the association between education and
occupation was stronger in France than in Germany. Notably, the cross-national
difference in educational margins is much more pronounced for men than for
women, which is a consequence of the fact that overall linkage differences are
much smaller for women than for men.
In 2010, the pattern has changed markedly: while the country differences
in linkage remain on the same order of magnitude, they are now to a much
lesser degree explained by compositional differences. Instead, about two-thirds
of the difference in M between Germany and France are now accounted for
by structural differences. Given the convergence in educational distributions
between Germany and France since 1970—mostly reflected by the expansion of
vocational training in France, with less change in Germany—it is not surprising
that the country differences in the educational marginal distributions contribute
much less to the overall country difference in M in 2010 than they did in 1970.
As in 2010, the linkage differences are smaller for women than for men.
Given the large size of the structural difference between France and Germany,
especially in 2010, it is interesting to study where these differences originate. The
two further decompositions reveal a considerable heterogeneity by educational
levels. First, the most striking result is that in 1970 graduates of ISCED
3ab_voc/3c/4 and 5b linked more strongly in France than in Germany; in
particular, the strength of the link between school and work was higher for
upper-secondary vocational graduates in France than in Germany. This again
confirms that, in contrast to what MSS argued, the French vocational system did
provide a close link between education and occupation. For tertiary educationTraining Regimes and Skill Formation in France and Germany 1133
the differences are smaller. We do find that in 2010 the average graduate with
a tertiary degree had higher linkage in Germany compared to France, but the
difference is small. The main reason why Germany links much stronger in
2010 than in 1970 stems from upper secondary vocational education (ISCED
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
categories 3ab_voc/3c/4), which contain both dual training and school-based
vocational education. For women, this category accounts for almost all of the
difference, while men also gained a small increase in linkage in higher education
(ISCED 5a/6).
The decomposition by field of study further shows that the French advantages
in 1970 were concentrated in services, business, engineering, and manufacturing.
Especially in engineering and manufacturing, men had higher structural linkage
than women, which likely reflects the focus of French vocational training on
these stereotypically male occupations. In 2010, higher linkage is visible across
most fields in Germany, where we can observe higher structural linkage for
men in fields such as science, maths, and computing, as well as engineering and
manufacturing. These trends likely reflect the continuing strong occupational
segregation by gender in both labor markets.
Within-country change over time
The results from the previous section indicate that there was substantial change
over time between the two countries, which implies that at least one of the
countries must itself have changed over time. Given that the structural differences
between Germany and France are much more pronounced in 2010 than 1970,
it is possible that Germany has increased its structural linkage, France has lost
some of its structural linkage, or a combination of these two processes. To answer
this question, we again use equation (3), but now apply this decomposition to
study change within each country over time (Figure 5).
As seen in Figure 1, observed total linkage has increased in both countries at
roughly comparable rates. However, the decomposition reveals that the reasons
for this increase are starkly different. In Germany, the counterfactual scenario
shows that the structure of linkage between educational programs and occupa-
tional destinations is almost equally strong in 1970 and 2010. The rise in linkage
strength is due to rapid growth in educational credentials that link more strongly
to the occupational structure. To be precise, 86 percent of the total difference
can be explained by changes in the educational and occupational distributions,
with the remaining 14 percent explained by increasing structural linkage. In
both countries, the occupational marginal components are higher for women
than for men, which is consistent with the idea of women “catching up” to the
occupational attainment of men.
In France, marginal and structural changes have different signs, and thus
partially offset each other. Due to educational expansion, the change in the
marginal educational and occupational distributions point to a large increase
in linkage. In fact, if in France only the educational distribution had changed
(without influencing structural linkage), France in 2010 would have almost the
same total linkage strength as Germany in 2010. However, a large and negative1134 Social Forces 99(3)
Figure 5. Decomposition of M within countries. Note: These results are also available as a table
in Appendix B
Downloaded from https://academic.oup.com/sf/article/99/3/1113/5847609 by Universiteit van Amsterdam user on 19 August 2021
component of structural change offsets a large part of the increase that would
be expected from changes in the marginals alone.
Much of the decrease in structural change is due to declines in vocational
education (3ab_voc/3c/4), and women have been much more affected by declines
in structural linkage than men. We again also find that most of the change
was concentrated in the fields of business, engineering, and manufacturing. In
Germany, the small increase in structural linkage is almost entirely due to the
male labor force, and especially concentrated in vocational education. If stronger
linkage provides benefits in the labor market, both countries have changed in
ways that have been unfavorable for women.
Changes in linkage for tertiary education have been small. This is another strong
indication that tertiary education is relatively unaffected from developments in
the vocational skill formation system. Both over time and between countries,
structural differences in tertiary education are small. In this sense, school-
to-work linkages in tertiary education are very similar between these two
countries and over time, while the differences for vocational education are more
pronounced.
Discussion: A historical perspective
Using data that span the 40 years from 1970 to 2010, we find that an in-depth
comparison of the French and German skill formation systems leads to a more
complex picture than commonly assumed. The French skill-formation systemYou can also read
